Use ORYKSA as the AI of your n8n automations

What this is for

Automations in n8n often need an AI model: to write a post from a new YouTube video, to classify a new row in Google Sheets, to summarise a file from Google Drive. Normally you would pay OpenAI or OpenRouter for that. With ORYKSA you do not need them: ORYKSA is the model, and it uses the interactions of your plan.

Your n8n keeps talking to YouTube, Instagram, LinkedIn, Drive or Sheets with your own accounts. ORYKSA only does the thinking and the writing.

What you need

• An ORYKSA account.
• An API key. Create it at developer.oryksa.com. It starts with oryk_live_. Keep it private, like a password.

Set it up in n8n

1. In n8n, open Credentials and create a new credential of type OpenAI.
2. In API Key, paste your ORYKSA key.
3. In Base URL, write https://api.oryksa.com/v1
4. Save. n8n tests the connection by itself.
5. In your flow, wherever there is an OpenAI Chat Model (in an AI Agent, a Basic LLM Chain, an Information Extractor, a Text Classifier), pick this credential and choose the model oryksa.
6. The Use Responses API option can be on or off: ORYKSA answers in both formats.

A flow you imported that says gpt-4o or another model name works too: ORYKSA answers in its place.

In Make, Zapier and other tools the idea is the same: where the tool asks for the OpenAI address and key, use the ORYKSA address and your ORYKSA key.

Two models

• oryksa: the general model. It does what your flow tells it and knows nothing about your account. Use this one for automations.
• oryksa-assistant: your own ORYKSA assistant, with the knowledge and the data of your business. It only answers.

What it uses

Each reply of the model uses one interaction of your plan. An AI Agent that calls tools gets several replies in a row, so it uses one interaction for each. When your interactions run out, the flow receives an error that says so. Nothing is charged beyond your plan.

Limits

Each request can carry up to 32 000 characters of text and each reply up to 2 000 tokens. For a long document, split it in parts in your flow.

If something does not work

• Invalid key: check that you pasted the whole key and that it was not revoked.
• The model does not appear: check the Base URL. It must end in /v1.
• Error with "Use Responses API" on: ORYKSA accepts text, function tools, JSON output and streaming in that format. It does not accept OpenAI's own tools (web search, file search, code interpreter) or stored conversations (previous_response_id): send the whole conversation in each request. The error message names the field.
• Request too large: the text is over 32 000 characters. Send less at a time.
• Limit reached: your plan has no interactions left in this period.